View
3
Download
0
Category
Preview:
Citation preview
arX
iv:0
806.
2804
v1 [
nucl
-th]
17
Jun
2008
Statistical Ensembles with Volume Fluctuations
Mark I. Gorenstein1
1Bogolyubov Institute for Theoretical Physics, Kiev, Ukraine
Abstract
The volume fluctuations in statistical mechanics are discussed. First, the volume fluctuations in
ensembles with a fixed external pressure, the so called pressure ensembles, are considered. Second,
a generalization of the pressure ensembles is suggested. Namely, the statistical ensembles with the
volume fluctuating according to externally given distributions are considered. Several examples and
possible applications in statistical models of hadron production are discussed.
PACS numbers: 24.10.Lx, 24.60.Ky, 25.75.-q
Keywords: statistical ensembles, pressure ensembles, volume fluctuations, particle number fluctuations
1
I. INTRODUCTION
Successful application of the statistical model to description of mean hadron multiplicities
in high energy collisions (see, e.g., recent papers [1] and references therein) has stimulated
investigations of properties of statistical ensembles. Whenever possible, one prefers to use the
grand canonical ensemble (GCE) due to its mathematical convenience. The canonical ensemble
(CE) [2] should be applied when the number of carriers of a conserved charges is small (of
the order of 1), such as strange hadrons [3], anti-baryons [4], or charmed hadrons [5]. The
micro-canonical ensemble (MCE) [6] has been used to describe small systems with fixed energy,
e.g. mean hadron multiplicities in proton-antiproton annihilation at rest. In all these cases,
calculations performed in different statistical ensembles yield different results. This happens
because the systems are ‘small’ and they are ‘far away’ from the thermodynamic limit (TL).
The mean multiplicity of hadrons in relativistic heavy ion collisions ranges from 102 to 104,
and mean multiplicities (of light hadrons) obtained within GCE, CE, and MCE approach each
other. One refers here to the thermodynamical equivalence of statistical ensembles and uses
the GCE for calculating the hadron yields.
Measurements of a hadron multiplicity distribution P (N) in interactions, including nucleus-
nucleus collisions, open a new field of applications of the statistical models. The particle
multiplicity fluctuations are usually quantified by the ratio of variance to mean value of a
multiplicity distribution P (N), the scaled variance, and are a subject of current experimental
activities. In statistical models there is a qualitative difference in the properties of mean
multiplicity and scaled variance of multiplicity distributions. It was recently found [7, 8, 9, 10,
11, 12, 13] that even in the TL corresponding results for the scaled variance are different in
different ensembles. Hence the equivalence of ensembles holds for mean values in the TL, but
does not extend to fluctuations.
A statistical system is characterized by the extensive quantities: volume V , energy E, and
conserved charge(s)1 Q. The MCE is defined by the postulate that all micro-states with given
V , E, and Q have equal probabilities of being realized. This is the basic postulate of the
statistical mechanics. The MCE partition function just calculates the number of microscopic
1 In statistical description of the hadron or quark-gluon systems, these conserved charges are usually the net
baryon number, strangeness, and electric charge. In non-relativistic statistical mechanics, the number of
particles plays the role of a conserved ‘charge’.
2
states with given fixed (V,E,Q) values. In the CE the energy exchange between the considered
system and ‘infinite thermal bath’ is assumed. Consequently, a new parameter, temperature
T is introduced. To define the GCE, one makes a similar construction for conserved charge Q.
An ‘infinite chemical bath’ and the chemical potential µ are introduced. The CE introduces
the energy fluctuations. In the GCE, there are additionally the charge fluctuations. The
MCE, CE, and GCE are the most familiar statistical ensembles. In several textbooks (see,
e.g., Ref. [14, 15]), the pressure (or isobaric) canonical ensemble has been also discussed. The
‘infinite bath of the fixed external pressure’ p0 is then introduced. This leads to the volume
fluctuations around the average value.
A more general concept of the statistical ensembles was suggested in Ref. [16]. The statistical
ensemble is defined by an externally given distribution of extensive quantities, Pα( ~A). The
construction of distribution of any property O in such an ensemble proceeds in two steps.
Firstly, the MCE O-distribution, Pmce(O; ~A), is calculated at fixed values of the extensive
quantities ~A = (V,E,Q). Secondly, this result is averaged over the external distribution Pα( ~A)
[16],
Pα(O) =
∫
d ~A Pα( ~A) Pmce(O; ~A) . (1)
The ensemble defined by Eq. (1), the α-ensemble, includes the standard statistical ensembles
as particular cases.
Recently, the micro-canonical ensemble with the volume fluctuations was introduced [17] for
modelling the hadron production in high energy interactions. An introduction of the volume
fluctuations was necessary in order to reproduce the KNO-scaling [18] of the hadron multiplicity
distribution observed in proton-proton and proton-antiproton collisions at high energies. The
volume fluctuations lead also to the power law shape of the single particle spectra at high
transverse momenta.
In the present paper the statistical ensembles with volume fluctuations are studied.
First, the pressure ensembles are considered. In general, there are 3 pairs2 of variables –
2 In the present study we do not discuss the role of the total 3-momentum. As shown in Ref. [13] the total
momentum conservation is not important in the TL for thermodynamical functions and fluctuations in the
full phase space. It may however influence the particle number fluctuations in the limited segments of the
phase space.
3
(V, p0), (E, T ), (Q, µ) – and, thus, the 8 statistical ensembles3 can be constructed. Among
these 8 ensembles there are 4 pressure ensembles: (p0, E,Q), (p0, T, Q), (p0, E, µ), and
(p0, T, µ). In addition to the pressure canonical ensemble known from the literature, three
other possibilities – pressure micro-canonical, pressure grand micro-canonical, and pressure
grand canonical ensembles – are constructed and studied. In Section II, non-relativistic sta-
tistical systems are discussed, whereas Section III presents the results for an ultra-relativistic
ideal gas.
In Section IV the concept of pressure ensembles is extended to the case of more general
volume fluctuations. Namely, the statistical ensembles with volume fluctuating according to
externally given distributions are introduced. The Summary presented in Section V closes the
paper.
II. NON-RELATIVISTIC BOLTZMANN GAS
A. Canonical and Micro-Canonical Ensembles
In this Section the system of non-relativistic Boltzmann particles is discussed. The (V, T,N)
Canonical Ensemble (CE) partition function of N particles reads [19]:
Zce(V, T,N) =1
N !
∫
dx1dp1
(2π)3. . .
dxNdpN
(2π)3exp
[
− E(x1, . . . ,xN ;p1, . . . ,pN)
T
]
, (2)
where T is the system temperature, the particle degeneracy factor is assumed to be equal to
1, and E is the microscopic N -particle energy usually presented as the sum of potential and
kinetic terms,
E(x1, . . . ,xN ;p1, . . . ,pN) = U(x1, . . . ,xN) +N∑
i=1
p2
i
2m, (3)
with m being the particle mass. For the N -particle energy given by (3) the integration over
momentum in Eq. (2) can be done explicitly,
Zce(V, T,N) =1
N !
(
mT
2π
)3N/2 ∫
V
dx1 . . . dxN exp
[
− U(x1, . . . ,xN)
T
]
. (4)
3 For several conserved charges {Qi} the number of possible ensembles is larger, as each charge can be treated
either canonically or grand canonically.
4
The particle coordinate x1, . . . ,xN are integrated over the system volume V . The CE thermo-
dynamical functions can be expressed in terms of the free energy,
F (V, T,N) = − T ln Zce(V, T,N) . (5)
The CE pressure, entropy, chemical potential, and average energy are:
p = −(
∂F
∂V
)
T,N
, S = −(
∂F
∂T
)
V,N
, µ =
(
∂F
∂N
)
T,V
, E = F + TS . (6)
For the non-interacting particles, the potential energy vanishes, U(x1, . . . ,xN) = 0, thus, the
ideal gas free energy is:
Fid(V, T,N) ∼= −NT − NT ln
[
V
N
(
mT
2π
)3/2]
, (7)
where we have assumed N ≫ 1 and, thus, lnN ! ∼= N lnN−N . The thermodynamical functions
of the ideal gas from Eqs. (6,7) read:
p =NT
V, S =
5
2N +N ln
[
V
N
(
mT
2π
)3/2]
, µ = −T ln
[
V
N
(
mT
2π
)3/2]
, E =3
2NT . (8)
The (V,E,N) Micro-Canonical Ensemble (MCE) partition function of N non-relativistic
Boltzmann particles reads:
Zmce(V,E,N) =1
N !
∫
dx1dp1
(2π)3. . .
dxNdpN
(2π)3δ
[
E − U(x1, . . . ,xN)−N∑
j=1
p2
j
2m
]
(9)
The MCE entropy is defined as,
S(V,E,N) = ln [E0 Zmce(V,E,N)] , (10)
where E0 is an arbitrary constant with a dimension of energy. For non-interacting particles,
U(x1, . . . ,xN) = 0, one finds,
Zmce(V,E,N) =V N(Em/2π)3N/2
E N ! Γ(3N/2), S(V,E,N) ∼= 5
2N +N ln
[
V
N
(
mE
3Nπ
)3/2]
. (11)
It has been assumed that N ≫ 1 and E ≫ E0, thus, Γ(3N/2) ∼= 3N/2 [ln(3N/2) − 1],
N ! ∼= N(lnN − 1), and ln(E0 · E3N/2−1) ∼= ln(E3N/2).
The MCE temperature, pressure, and chemical potential are the following:
1
T=
(
∂S
∂E
)
V,N
,p
T=
(
∂S
∂V
)
E,N
,µ
T= −
(
∂S
∂N
)
V,E
. (12)
5
For non-interacting particles they read,
1
T=
3
2
N
E,
p
T=
N
V,
µ
T= − ln
[
V
N
(
mE
3Nπ
)3/2]
. (13)
If E = E, the ideal gas system of N particles in the volume V has the same temperature,
pressure, and entropy in the MCE (11-13) and in the CE (8). This means the thermodynamical
equivalence of CE and MCE at N ≫ 1.
B. Pressure Canonical and Pressure Micro-Canonical Ensembles
The particle coordinates and momenta for the considered system are denoted as x1, . . . , xN
and p1, . . . , pN , respectively. The E given by Eq. (3) denotes the system energy, and V is the
system volume. The ‘thermostat’ is now introduced with corresponding particle coordinates,
X1, . . . , XNT, and momenta, P1, . . . , PNT
, and with ET and VT being the thermostat
energy and volume, respectively. The system plus thermostat is described by the MCE, i.e.
the total energy and the total volume are assumed to be fixed: E + ET = E∗ = const ,
V +VT = V ∗ = const . The probability distribution of system particle coordinates and momenta
is proportional to,
fV (x1, . . . , xN ; p1, . . . , pN ) ∝∫ NT∏
i=1
d3Xid3Pi δ
(
E +
NT∑
j=1
P2
j
2M− E∗
)
, (14)
where the thermostat particle coordinates X1, . . . ,XNTare integrated over the thermostat
volume VT , and the system particle coordinates x1, . . . ,xN are contained in the volume V . The
particles in the thermostat are assumed to be non-interacting. Thus, ET =∑NT
j=1P2
j/(2M),
with M being the particle mass. The momentum integration in Eq. (14) gives,
∫
d3P1 . . . d3PNT
δ
(
E +
NT∑
j=1
P2
j
2M− E∗
)
∝ (E∗ − E)3NT /2 −1 ∝(
1− E
E∗
)3NT /2 −1
∼=(
1 − E
ET
)3NT /2 −1
∼=(
1 − E
3NTT/2
)3NT /2
∼= exp(− E/T ) , (15)
where it has been assumed, NT → ∞, E/ET → 0, and the ideal gas equation for the thermostat
energy, ET = 3TNT/2, has been used. The integration over thermostat particle coordinates Xi
6
in Eq. (14) gives the factor depending on the system volume V ,
∫
VT
d3X1 . . . d3XNT
= V NT
T = (V ∗ − V )NT ∝(
1 − V
V ∗
)NT
∼=(
1 − V
VT
)NT
=
(
1 − V
(TNT/p0)
)NT
∼= exp
(
− p0 V
T
)
, (16)
where it has been assumed, VT → ∞, NT → ∞, V/VT → 0, and the ideal gas equation for
the thermostat pressure, p0 = TNT/VT , has been used.
One obtains from Eqs. (14-16),
fV (x1, . . . ,xN , p1, . . . ,pN) =1
Zpce(p0, T, N)
1
N !exp
(
− p0V + E
T
)
, (17)
where the system energy E in (17) depends on particle coordinates and momenta according to
Eq. (3). The function Zpce(p0, T, N) in Eq. (17) is the (p0, T, N) Pressure Canonical Ensemble
(PCE) partition function. It is defined by the normalization condition,
∫
∞
0
dV
∫ N∏
i=1
[
d3xid3pi
(2π)3
]
fV (x1, . . . ,xN ; p1, . . . ,pN) = 1 , (18)
and it equals to,
Zpce(p0, T, N) =
∫
∞
0
dV exp
(
− p0 V
T
)
Zce(V, T,N) , (19)
where Zce(V, T,N) is the CE partition function given by Eq. (2).
Using Zce = exp(−F/T ), the probability volume distribution in the PCE can be presented
as:
Wpce(V ) =1
Zpce(p0, T, N)exp
[
− F (V, T,N) + p0V
T
]
. (20)
First, one finds the maximum of Wpce(V ), which defines the most probable value of the volume,
V = V0,[
∂F (V, T,N)
∂V
]
V=V0
+ p0 = 0 . (21)
By definition, − ∂F/∂V is the CE pressure (8). Thus, the PCE equation of state (21) has clear
physical meaning: the internal pressure p for the most probable volume V0 equals to the fixed
external pressure p0,
p(V0, N, T ) = p0 . (22)
7
In the TL the volume distribution (20) can be approximated as,
Wpce(V ) ∼= 1
Zpce(p0, T, N)exp
[
− F (V0, T, N) + p0V0
T
]
(23)
× exp
[
− 1
2T
(
∂2F (V, T,N)
∂V 2
)
V=V0
(V − V0)2
]
≡ (2πωV V0)−1/2 exp
[
− (V − V0)2
2ωV V0
]
.
Thus, the most probable volume, V0, and the average volume V are equal to each other in the
TL,
V ≡∫
∞
0
dV V Wpce(V ) ∼= V0 , (24)
and ωV introduced in Eq. (23) defines the scaled variance of the volume fluctuations,
ωV ≡ V 2 − V2
V=
T
V0
[
∂2F (V, T,N)
∂V 2
]
−1
V=V0
= − T
V0
[
∂p(V, T,N)
∂V
]
−1
V=V0
. (25)
For the ideal gas pressure (8) one gets:
V ∼= V0 =NT
p0, ωV
∼= T
p0. (26)
The Eqs. (23-25) are valid if (∂p/∂V ) is negative. This is the case for the ‘normal’ equation of
state, but is not valid for either a 1st order phase transition or critical point. Note that the TL
N, V0 → ∞, with N/V0 being finite, is assumed to make clear notions of the phase transition
or critical point.
1st Order Phase Transition. In the casei of the 1st order phase transition ∂p/∂V = 0 [19] for
V1 < V < V2 in the mixed phase. According to Eq. (20) this leads to Wpce(V ) = const ∼=(V2 − V1)
−1 for V1 < V < V2. Introducing the notation, V2 = γV1, with γ = const > 1, one
finds at V1 → ∞,
V ∼=∫ V2
V1
V Wpce(V )dV =γ + 1
2V1 , (27)
ωV∼= 1
V
[∫ V2
V1
V 2 Wpce(V )dV − V2
]
=1
3
(
γ − 1
γ + 1
)2
V . (28)
Thus, the volume fluctuations are anomalously large, ωV ∝ V → ∞.
Critical Point. At the critical point it follows [19],
∂p
∂V=
∂2p
∂V 2= 0 . (29)
8
In this case, Eq. (20) takes the form,
Wpce(V ) ∼= 1
Zpce(p0, T, N)exp
[
− F (V0, T, N) + p0V0
T
]
(30)
× exp
[
− 1
T
1
4!
(
∂4F (V, T,N)
∂V 4
)
V=V0
(V − V0)4
]
≡ 2 A1/40
Γ(1/4)exp
[
− A0 (V − V0)4]
,
where A0 ≡ −(24T )−1(∂3p/∂V 3)V=V0. The volume distribution (30) leads to the scaled variance
for the volume fluctuations,
ωV =V 2 − V
2
V∼= 1
V0
∫
∞
0
dV (V − V0)2 Wpce(V ) =
1
V0
Γ(3/4)
Γ(1/4)A
−1/20
, (31)
where the Gamma functions in Eqs. (30-31) are Γ(1/4) ∼= 3.626 and Γ(3/4) ∼= 1.225. Finally,
one finds for (31),
ωV∼= 1.656 T 1/2
V0
[
−(
∂3p
∂V 3
)
V=V0
]
−1/2
. (32)
In order to estimate the scaled variance (32) of the volume fluctuations at the critical point we
use the van der Waals (VdW) equation of state [19],
(
p + aN2
V 2
)
(V − b N) = NT . (33)
The critical point is defined by the following equations
∂p
∂V= − NT
(V −Nb)2+
2N2a
V 3= 0 , (34)
∂2p
∂V 2=
2NT
(V −Nb)3− 6N2a
V 4= 0 . (35)
They give,
Tcr =8
27
a
b, Vcr = 3Nb , pcr =
1
27
a
b2. (36)
At the critical point one finds,
(
∂3p
∂V 3
)
cr
=
[
− 6NT
(V −Nb)4+
24N2a
V 5
]
cr
= − a
92b51
N3. (37)
Thus, for T = Tcr, p0 = pcr, and V ∼= V0 = Vcr, and Eqs. (32,37) give,
ωV∼= 1.656 T 1/2
Vcr
[
−(
∂3p
∂V 3
)
V=Vcr
]
−1/2
∼= 2.703 b√N . (38)
9
The volume fluctuations at the critical point are anomalously large, ωV ∝ N1/2 ∝ V1/2 → ∞.
Similarly to the PCE we introduce now the (s0, E,N) Pressure Micro-Canonical ensemble
(PMC),
Zpmc(s0, E,N) =
∫
∞
0
dV exp (−s0V )Zmce(V,E,N) , (39)
where the parameter s0 = p0/T0 is defined by the external conditions of the thermostat pressure
p0 and the thermostat temperature T0. The micro-canonical energy E is assumed to be fixed.
Thus, there is no energy exchange between the ‘thermostat’ and considered system, and the
internal MCE temperature T (12) may differ from T0. Using Eq. (10), the volume distribution
in the PMC can be presented in the form,
Wpmc(V ) =1
Zpmc(s0, E,N)exp [− s0 V + S(V,E,N)] (40)
∼= 1
Zpmc(s0, E,N)exp
[
− s0 V0 + S(V0, E,N) +1
2
(
∂2S
∂V 2
)
V=V0
(V − V0)2
]
.
The most probable (and average) volume V0 is defined by the condition,
− s0 +
(
∂S(V,E,N)
∂V
)
V=V0
= 0 . (41)
The Eq. (41) corresponds to,
p
T=
p0T0
, (42)
which is similar to Eq. (22) in the PCE. The scaled variance of the volume fluctuations ωV in
the PMC equals to:
ωV∼= −
[
V0 (∂2S/∂V 2)V=V0
]
−1
= − T
V0
[
∂p(V,E,N)
∂V
]
−1
V=V0
, (43)
which is again similar to Eq. (25) in the PCE. Thus, the average volume V0 and scaled variance
of the volume fluctuations ωV are identical in the PCE and PMC in the TL. For the ideal gas
one finds,
〈V 〉pmc∼= V0 =
N
s0, ωV =
1
s0, (44)
which coincide with (26) in the PCE.
10
C. Grand Canonical Ensemble
The (V, T, µ) Grand Canonical Ensemble (GCE) partition function of non-relativistic non-
interacting Boltzmann particles reads:
Zgce(V, T, µ) =∞∑
N=0
exp
(
µ N
T
)
Zce(V, T,N) , (45)
where Zce(V, T,N) is given by Eq. (2) and µ is the chemical potential. The GCE pressure is
calculated as,
p(T, µ) =T
Vln [Zgce(V, T, µ)] , (46)
and it does not depend on the volume V in the TL. The CE, MCE, and GCE are thermo-
dynamically equivalent at V → ∞. Note that the GCE is the most convenient one from the
technical point of view.
There is, however, an evident problem in the formulation of the (p0, T, µ) Pressure Grand
Canonical ensemble (PGC). The PGC partition function is obtained by extending Eq. (19),
Zpgc(p0, T, µ) =
∫
∞
0
dV exp
(
− p0 V
T
)
Zgce(V, T, µ) =T
p0 − p(T, µ), (47)
where p(T, µ) is the GCE pressure (46). The (p0, T, µ)-ensemble has a unique property. Among
8 possible ensembles this is the only one where the system description includes only intensive
quantites, p0, T and µ. For p(T, µ) = p0, the system volume is undefined. In the domain
p(T, µ) ≥ p0, the PGC partition function does not exist as the integral over the volume in
Eq. (47) diverges. For p(T, µ) < p0, the volume distribution has the form:
Wpgc(V ) =1
Zpgc(p0, T, µ)exp
[
− V
(
p0 − p
T
)]
. (48)
The most probable volume V0 equals to zero, but the average volume is:
V =
∫
∞
0
dV V Wpgc(V ) =T
p0 − p(T, µ). (49)
This special ensemble will be discussed further in details for the ultra-relativistic gas in the
next section.
11
III. ULTRA-RELATIVISTIC GAS
In this section several examples of the pressure ensembles for the ultra-relativistic (m = 0)
ideal gas of Boltzmann particles are considered. For simplicity only statistical systems without
conserved charges are discussed. Thus, the number of particles is not restricted and chemical
potential equals to zero. Furthermore, the Boltzmann statistics is used and the degeneracy
factor is assumed to be one.
A. Grand Canonical and Grand Micro-Canonical Ensembles
The (V, T ) GCE4 partition function of massless non-interacting neutral Boltzmann particles
reads:
Zgce(V, T ) =∞∑
N=0
1
N !
(
V
2π2
)N ∫ ∞
0
N∏
i=1
p2idpi exp(
−piT
)
=∞∑
N=0
1
N !
(
V T 3
π2
)N
= exp(N) , (50)
where N ≡ 〈N〉gce = V T 3/π2 is the GCE average number of particles. The GCE system
pressure and average energy are:
p =T
VlnZgce =
T 4
π2= T n(T ) , (51)
〈E〉gce ≡ E = T 2∂ lnZgce
∂T=
3
π2V T 4 = ε(T ) V , (52)
where n(T ) = N/V = T 3/π2 and ε(T ) = 3T 4/π2 are the particle number density and energy
density, respectively. The GCE multiplicity distribution has the Poisson form,
Pgce(N ;V, T ) =N
N
N !exp
(
− N)
, (53)
and the scaled variance of particle number distribution (53) equals to:
ωgce ≡〈N2〉gce − 〈N〉2gce
〈N〉gce= 1 . (54)
4 The chemical potential connected to the number of particles equals to zero.
12
The (V,E) Grand Micro-Canonical ensemble5 (GMC) partition function is [8]:
Zgmc(V,E) =
∞∑
N=1
1
N !
(
V
2π2
)N ∫
∞
0
N∏
i=1
p2idpi δ
(
E −N∑
j=1
pj
)
≡∞∑
N=1
WN(V,E) =1
E
∞∑
N=1
AN
N ! (3N − 1)!=
A
2E0F3
(
;4
3,5
3, 2;
A
27
)
, (55)
where 0F3 is the generalized hyper-geometric function [21], and
A ≡ V E3
π2. (56)
The GMC particle number distribution function equals to:
Pgmc(N ;V,E) ≡ WN(V,E)
Zgmc(V,E)=
1
Zgmc(V,E)
AN
E N ! (3N − 1)!. (57)
It is defined for N ≥ 1. The average number of particles in the GMC equals to [8]:
〈N〉gmc∼=(
A
27
)1/4
. (58)
In the large volume limit the mean multiplicities in the GME and the GCE are equal, 〈N〉gmc =
N providing E = E and the GMC and the GCE volumes are equal. The temperature and the
pressure in the GME are equal to:
T =E
3N=
(
π2E
3V
)1/4
, p =NT
V=
T 4
π2, (59)
and they coincide with the corresponding quantities (51-52) in the GCE.
For N ≫ 1 the particle number distribution in the GME (57) can be approximated by the
Gaussian one:
Pgmc(N ;V,E) ∼=(
2π ωgce N)
−1/2exp
[
−(
N −N)2
2 ωgmc N
]
, (60)
with the GME scaled variance
ωgmc =〈N2〉gmc − 〈N〉2gmc
〈N〉gmc=
1
4. (61)
The Poisson distribution Pgce(N) (53) for N ≫ 1 can be also approximated by the Gauss
distribution (60), but with ωgce = 1. In the TL the particle number distributions in the
GME and GCE have both the Gauss form [12]. The average number of particles is the same,
〈N〉gmc∼= N , but the scaled variance is different, ωgce = 1 and ωgmc = 1/4.
5 The energy is fixed, but the number of particles is not. The chemical potential connected to the number of
particles equals to zero. Thus, this ensemble is named the Grand Micro-Canonical Ensemble [9].
13
B. Pressure Grand Canonical Ensemble
The (p0, T ) Pressure Grand Canonical ensemble (PGC) is defined as the ensemble with the
fixed external pressure p0, temperature T , and the chemical potential connected to the number
of particles equals to zero. The PGC partition function equal to [20]:
Zpgc(p0, T ) ≡∫
∞
0
dV exp
(
− p0V
T
)
Zgce(V, T ) =T
p0 − p(T ), (62)
where the relation (51) between p(T ) and Zgce has been used. The value of p0 has a physical
meaning of the external pressure. A convergence of the integral over the volume in Eq. (62)
requires the inequality,
p(T ) =1
π2T 4 < p0 . (63)
Thus, at each fixed value of p0, there is a ‘limiting temperature’ T ∗ in the PGC:
T < T ∗ =(
π2 p0)1/4
. (64)
The probability volume distribution in the PGC is:
Wpgc(V ) =1
Zpgc(p0, T )exp
[
− V
(
p0 − p(T )
T
)]
. (65)
It gives the most probable volume equal to zero, V0 = 0, and the average value,
〈V 〉pgc ≡ V =
∫
∞
0
dV V Wpgc(V ) = − T∂ lnZpgc(p0, T )
∂p0=
T
p0 − p(T ), (66)
which is not equal to zero and may even go to infinity at T → T ∗. Using Eq. (66), the volume
distribution (65) can be written as,
Wpgc(V ) = V−1
exp(
− V/V)
. (67)
For the mean values of energy and particle number in the PGC one finds:
〈E〉pgc =
∫
∞
0
dV Wpgc(V ) 〈E〉gce =3
π2T 4 V = ε(T ) V , (68)
〈N〉pgc =
∫
∞
0
dV Wpgc(V ) 〈N〉gce =1
π2T 3 V = n(T ) V , (69)
where n(T ) and ε(T ) are the particle number density and energy density of the GCE given by
Eq. (51) and Eq. (52), respectively.
14
The independent variables are (V, T ) in the GCE and (p0, T ) in the PGC. If the GCE
volume is chosen to be equal to the average volume of the PGC, then Eqs. (68) and (69) give:
〈E〉pgc = E and 〈N〉pgc = N , i.e. the GCE and PGC are thermodynamically equivalent. The
fluctuations of E and N are however different in these two ensembles. Calculating,
〈V 2〉pgc =
∫
∞
0
dV V 2 Wpgc(V ) =T 2
Zpgc
∂2Zpgc
∂p20
= 2
[
T
p0 − p(T )
]2
= 2V2
, (70)
one finds,
ωV =〈V 2〉pgc − 〈V 〉2pgc
〈V 〉pgc= V . (71)
Thus, in the TL limit V → ∞ the volume fluctuations become anomalously large. This, in turn,
leads to anomalous energy and particle number fluctuations. The particle number distribution
in the PGC has the form of the geometrical distribution:
Ppgc(N ; p0, T ) =
∫
∞
0
dV Wpgc(V ) Pgce(N, V, T ) = (1− η) ηN , (72)
where Pgce(N, V, T ) has been taken from Eq. (53), and η ≡ (T/T ∗)4 < 1 . The most probable
number of particles is N = 0, whereas the average value (69) is larger than zero. It equals to
〈N〉pgc = η(1 − η)−1 and may even go to infinity at η → 1. This happens if T → T ∗. From
Eq. (72) it follows:
ωpgc =〈N2〉pgc − 〈N〉2pgc
〈N〉pgc= 1 +
η
1− η= 1 + n(T ) V . (73)
The first term in the r.h.s. of Eq. (73) corresponds to the Poisson fluctuations (54) of the GCE
at fixed volume, whereas the second term comes from the volume fluctuations at fixed particle
number density. The multiplicity distribution (72) can be rewritten as,
Ppgc(N ; p0, T ) ≡ Ppgc(N ;N) =1
N + 1exp
[
− N ln
(
1 +1
N
)]
. (74)
For N ≫ 1 the distribution Ppgc approches:
Ppgc(N ;N) ∼= 1
Nexp
(
− N
N
)
. (75)
The particle number distribution Ppgc(N) (75) satisfies the so called KNO-scaling [18].
15
C. Pressure Grand Micro-Canonical Ensemble
The (s0, E) Pressure Grand Micro-canonical ensemble (PGM) is defined as the statistical
ensemble with fixed energy E and fixed external parameter s0 = p0/T0. As before, the chemical
potential connected to the number of particles equals to zero. The PGM partition function is
equal to:
Zpgm(s0, E) ≡∫
∞
0
dV exp(− s0V ) Zgmc(V,E) (76)
=1
E
∞∑
N=1
(
E3
π2
)N1
N ! (3N − 1)!
∫
∞
0
dV exp(− s0V ) V N =1
s0 E
∞∑
N=1
BN
(3N − 1)!,
where
B ≡ E3
s0 π2. (77)
Using Eq. [22],
∞∑
k=0
x3k
(3k)!=
1
3
[
exp(x) + 2 exp(−x/2) cos
(√3
2x
)]
, (78)
one finds at B ≫ 1,
ln[Zpgm(s0, E)] ∼= B1/3 . (79)
In the TL, the volume distribution in the PGM can be found from (76) using the asymptotic
behavior of 0F3 function [21],
ln[Zgmc(V,E)] = ln
[
A
2E0F3
(
;4
3,5
3, 2;
A
27
)]
∼= 4
(
V E3
27π2
)1/4
. (80)
The volume distribution in the PGM is then proportional to,
Wpgm(V ) ∝ exp
[
4
(
V E3
27π2
)1/4
− s0V
]
≡ exp[φ(V )]
∼= exp
[
φ(V0) +1
2
(
∂2φ
∂V 2
)
V=V0
(V − V0)2
]
. (81)
The most probable volume V0 in the PGM is defined by the condition,
(
∂φ
∂V
)
V=V0
=
(
E3
27π2V 3
0
)1/4
− s0 = 0 . (82)
16
This gives,
V0 =E
3π2/3s4/30
. (83)
One also finds,
(
∂2φ
∂V 2
)
V=V0
= − 3s04V0
. (84)
Thus, the volume distribution in the PGM can be approximated as:
Wpgm(V ) ∼= (2π ωV V0)−1/2 exp
[
− (V − V0)2
2 ωV V0
]
, (85)
where ωV = 4/(3s0) is the scaled variance of the volume fluctuations in the PGM. The average
volume and its fluctuations in the PGM can be also calculated in terms of the partition function
Zpgm(s0, E) using Eqs. (76-79),
〈V 〉pgm =
∫
∞
0
dV Wpgm(V ) V = − ∂ ln Zpgm
∂s0∼= E
3π2/3s4/30
, (86)
〈V 2〉pgm =
∫
∞
0
dV Wpgm(V ) V 2 =1
Zpgm
∂2Zpgm
∂s20
∼= E2
9π4/3s8/30
, (87)
ωV =〈V 2〉pgm − 〈V 〉2pgm
〈V 〉pgm∼= 4
3s0. (88)
The condition (83) can be written as,
s0 =1
π2T 3 =
p
T, (89)
and it means that the internal pressure p equals to s0T . The most probable volume V0 (83) and
average volume 〈V 〉pgm (86) are then equal to each other, and both are equal to Eπ2/(3T 4). This
corresponds to the fixed volume in the MCE with fixed energy E and the MCE temperature
T . The volume distribution in the PGM is therefore different from that in the PGC (67). In
contrast to the PGC, there is an extensive variable, the energy E, in the PMG. This leads to the
system average volume proportional to the energy and given by Eq. (83). The internal pressure
equals to p = s0T (89). As a result, the volume fluctuations in the PGM are Gaussian (85)
with the finite scaled variance (88). Using Eq. (89) the scaled variance (88) of the volume
fluctuations can be expressed in terms of the MCE temperature, ωV = 4π2/(3T 3). One also
17
finds:
〈N〉pgm =1
ZpgmB
∂Zpgm
∂B∼= 1
3B1/3 , (90)
〈N2〉pgm =1
ZpgmB
∂
∂BB
∂Zpgm
∂B∼= 1
9B1/3 +
1
9B2/3, (91)
ωpgm =〈N2〉pgm − 〈N〉2pgm
〈N〉pgm∼= 1
3. (92)
In the TL, the multiplicity distribution in the PGM can be approximated as:
Ppgm(N ; s0, E) ∼=(
2π ωpgm N)
−1/2exp
[
−(
N −N)2
2 ωpgm N
]
, (93)
where N = B1/3/3 ∼= 〈N〉pgm and ωpgm = 1/3 . The scaled variance in the PGM can be
presented in the TL as the following,
ωpgm = ωmce +1
16ωV n =
1
4+
1
12=
1
3, (94)
where n = T 3/π2 is the MCE particle number density. The first term in the r.h.s. of Eq. (94),
ωmce = 1/4 , is due to the particle number fluctuations in the MCE with fixed volume, and the
second term is the contribution due to the volume fluctuations.
IV. ENSEMBLES WITH EXTERNAL VOLUME FLUCTUATIONS
The multiplicity distributions P (N) in relativistic gases [7, 8, 9, 10, 11, 12, 13] are sensitive
to conservation laws obeyed by the system, and therefore to fluctuations of extensive quantities
E and Q. The examples considered in the previous section demonstrate that the volume fluctu-
ations also influence the particle number fluctuations. Thus, for the calculation of multiplicity
distributions, the choice of the statistical ensemble is then not a matter of convenience, but a
physical question. On the other hand, the fluctuations of extensive quantities ~A ≡ (V,E,Q)
depend not on the system’s physical properties, but rather on external conditions. One can
imagine a huge variety of these conditions, thus, 8 statistical ensembles discussed in the previous
sections are only some special examples.
A more general concept of the statistical ensembles based on Eq. (1) was recently suggested
in Ref. [16]. The system volume may exhibit fluctuations described by the externally given
distribution. When V is the only fluctuating variable, Eq. (1) is reduced to
Pα(O) =
∫
dV Pα(E, V,Q) Pmce(O;E, V,Q) , (95)
18
where Pα(V ) is externally given volume distribution.
The effect of volume fluctuations on the particle number fluctuations is calculated for
the system of non-interacting massless Boltzmann particles with zero chemical potential. At
fixed volume the system is treated within the GMC, and the particle number distribution is
Pgmc(N ;V,E) (57).
In the first example, the volume distribution is assumed to be:
Pα(V ) =(
2πωV a2V V)
−1/2exp
[
−(
V − V)2
2ωV a2V V
]
, (96)
where ωV = 4π2/(3T 3) coincides with (88) in the PGM. The choice of Pα(V ) results in a simple
correspondence to the GMC and PGM in the TL. In Eq. (96), aV is a dimensionless tuneable
parameter which determines the width of the distribution. In the limit aV → 0, Eq. (96)
becomes a Dirac δ-function, δ(V − V ). This corresponds to the GMC. For aV = 1, Eq. (96)
results in the PGM volume fluctuations (85) in the TL. The particle number distribution reads,
Pα(N) =
∫
∞
0
dV Pα(V ) Pgmc(N ;V,E) . (97)
A substitution of Pα(V ) in Eq. (97) by the distribution (96) results in Pα(N) in the TL given
by the Gaussian,
Pα(N) ∼=(
2πωαN)
−1/2exp
[
−(
N − N)2
2ωα N
]
, (98)
where the average number of particles N is defined by the energy and average volume,
N = [V E3/(27π2)]1/4, and the scaled variance of the particle number distribution (98) equals,
ωα =1
4+
1
12a2V . (99)
The first term in the r.h.s. of Eq. (99) equals to the GMC scaled variance at fixed E and V , the
second term is due to the volume fluctuations. As it can be expected, ωα = ωgmc for aV = 0,
and ωα = ωpgm for aV = 1. It also follows, ωgmc < ωα < ωpgm for 0 < aV < 1, and ωα > ωpgm for
aV > 1. The α-ensemble defined by Eqs. (97, 96) presents an extension of the GMC (aV = 0)
and PGM (aV = 1) to a more general volume distribution.
In the second example both E and V are assumed to fluctuate. Equation (97) should be
then extended as:
Pα(N) =
∫
dEdV Pα(E, V ) Pgmc(N ;E, V ) . (100)
19
First, uncorrelated volume and energy distributions are considered:
Pα(E, V ) = P1(V )× P2(E) , (101)
where P1(V ) is given by Eq. (96) and P2(E) is taken in the following form,
P2(E) =(
2πωE a2EE)
−1/2exp
[
−(
E − E)2
2ωE a2E E
]
, (102)
where E = 3T 4V /π2 and ωE = 4T . A substitution of Pα(E, V ) in Eq. (100) by the above distri-
butions results in Pα(N) in the TL given again by the Gaussian (98) with the average number
of particles N defined by the average energy and average volume, N = [V E3
/(27π2)]1/4, and
the scaled variance equal to:
ωα =1
4+
3
4a2E +
1
12a2V . (103)
The first term in the r.h.s. of Eq. (103) is the MCE scaled variance at fixed E and V , the second
term is due to energy fluctuations, and the third one comes from volume fluctuations. The
results of the GMC, PGM, and GCE for the scaled variance of the particle number distribution
can be obtained from Eq. (103) at aV = aE = 0; aV = 1, aE = 0; and aV = 0, aE = 1,
respectively
Examples of correlated V and E distributions can be constructed as follows. One assumes
the factorized form (101) with Eq. (102) for the energy distribution, but with E = 3T 4V/π2
and ωE = 4T , i.e. the average energy E depends on the volume V . Assuming,
P1(V ) =1
Vexp(− V/V ) , (104)
with V given by Eq. (66), from Eqs. (101) and (102) with aE = 1, the results for the PGC are
obtained.
The relation, E = 3T 4V/π2, can be used together with any form of the volume distribution
P1(V ). This yields a generalization of the GCE to the systems with externally given volume
fluctuations. The energy fluctuations at fixed V can be also selected in accordance with the
physics requirements.
As the third example of the statistical ensembles with volume fluctuations one refers to the
recent paper [17] where the micro-canonical ensemble with the scaling volume fluctuations for
the ultra-relativistic ideal gas has been considered. The volume fluctuations were assumed to
20
have the specific scaling properties. They were chosen to describe the KNO scaling [18] of
the particle multiplicity distributions measured in proton-proton collisions at high energies.
A striking feature of the model is power law form of the single momentum spectrum at high
momenta, instead of the exponential Boltzmann distribution in the systems with fixed volume.
V. SUMMARY
In this paper the volume fluctuations in the statistical mechanics have been studied. The
statistical ensembles with fixed external pressure were considered. Statistical systems of classi-
cal non-relativistic particles and non-interacting massless particles were discussed. The volume
fluctuations in the pressure canonical ensemble become anomalously large in the thermody-
namic limit for the first order phase transition, ωV ∝ N , or at the critical point, ω ∝ N1/2.
Another type of anomalous volume fluctuations takes place for the pressure grand canonical
ensemble. In this special ensemble, all thermodynamical variables are the intensive quanti-
ties, (p0, T, µ). In the thermodynamical limit the mean particle multiplicity obtained within
the considered ensembles with volume fluctuations is equal to the mean calculated within the
micro-canonical, canonical, and grand canonical ensembles. This is not valid, however, for the
scaled variance. The influence of the volume fluctuations in the pressure ensembles on particle
number fluctuations have been discussed for ultra-relativistic ideal gas.
Following Ref. [16] the ensembles with volume fluctuations determined by the externally
given distribution function were introduced. Multiplicity fluctuations are sensitive to the vol-
ume fluctuations. The volume fluctuations may also influence the behavior of the particle
momentum spectra [17]. Thus, we believe that the concept of statistical ensembles with fluc-
tuating extensive quantities, in particular, with fluctuating volume of the statistical system,
may be appropriate for the statistical description of hadron production in relativistic collisions.
It may be also useful for other physical systems. In fact, in all cases when the equilibrium
statistical mechanics is used to calculate the fluctuations of the system properties.
Acknowledgments
We would like to thank D.V. Anchishkin, V.V. Begun, M. Gazdzicki, W. Greiner, M. Hauer,
B.I. Lev, I.N. Mishustin, O.N. Moroz, and Yu.M. Sinyukov for numerous discussions. This
21
work was in part supported by the Program of Fundamental Researches of the Department of
Physics and Astronomy of NAS, Ukraine.
[1] J. Cleymans, H. Oeschler, K. Redlich, and S. Wheaton, Phys. Rev. C 73, 034905 (2006); F. Becat-
tini, J. Manninen, and M. Gazdzicki, ibid. 73, 044905 (2006); A. Andronic, P. Braun-Munzinger,
and J. Stachel, Nucl. Phys. A 772, 167 (2006).
[2] F. Becattini, Z. Phys. C 69, 485 (1996); F. Becattini and U. Heinz, Z. Phys. C 76, 269 (1997).
[3] J. Cleymans, K. Redlich, and E. Suhonen, Z. Phys. C 51, 137 (1991).
[4] M.I. Gorenstein, M. Gazdzicki, and W. Greiner, Phys. Lett. B 483, 60 (2000).
[5] M.I. Gorenstein, A.P. Kostyuk, H. Stocker, and W. Greiner, Phys. Lett. B 509, 277 (2001).
[6] F. Becattini and L. Ferroni, Eur. Phys. J. C 35, 243 (2004); 38, 225 (2004); V.V. Begun, L. Fer-
roni, M.I. Gorenstein, M. Gazdzicki, F. Becattini, J. Phys. G 32, 1003 (2006); F. Becattini and
L. Ferroni, Eur. Phys. J. C 51, 899 (2007); 52, 597 (2007).
[7] V.V. Begun, M. Gazdzicki, M.I. Gorenstein, and O.S. Zozulya, Phys. Rev. C 70, 034901 (2004);
V.V. Begun, M.I. Gorenstein, and O.S. Zozulya, Phys. Rev. C 72, 014902 (2005); A. Keranen,
F. Becattini, V.V. Begun, M.I. Gorenstein, and O.S. Zozulya, J. Phys. G 31, S1095 (2005);
F. Becattini, A. Keranen, L. Ferroni, and T. Gabbriellini, Phys. Rev. C 72, 064904 (2005);
J. Cleymans, K. Redlich, and L. Turko, Phys. Rev. C 71, 047902 (2005); J. Phys. G 31, 1421
(2005); V.V. Begun and M.I. Gorenstein, Phys. Rev. C 73, 054904 (2006).
[8] V.V. Begun, M.I. Gorenstein, A.P. Kostyuk, and O.S. Zozulya, Phys. Rev. C 71, 054904 (2005).
[9] V.V. Begun, M.I. Gorenstein, A.P. Kostyuk, and O.S. Zozulya, J. Phys. G 32, 935 (2006).
[10] V.V. Begun, M.I. Gorenstein, M. Hauer, V.P. Konchakovski, and O.S. Zozulya, Phys. Rev. C 74,
044903 (2006);
[11] V.V. Begun, M. Gazdzicki, M.I. Gorenstein, M. Hauer, B. Lungwitz, and V.P. Konchakovski,
Phys. Rev. C 76, 024902 (2007).
[12] M. Hauer, V.V. Begun, and M.I. Gorenstein, arXiv:0706.3290 [nucl-th].
[13] M. Hauer, arXiv:0710.3938 [nucl-th].
[14] Yu.B. Rumer and M. Sh. Rivkin, Thermodynamics, Statistical Physics, and Kinetics (Nauka,
Moscow, 1972) [in Russian].
22
[15] K.B. Tolpygo, Thermodynamics and Statistical Physics (Kiev University, 1966) [in Russian].
[16] M.I Gorenstein and M. Hauer, arXiv:08014219 [nucl-th].
[17] V.V. Begun, M. Gazdzicki, and M.I. Gorenstein, arXiv:0804.0075 [hep-ph].
[18] Z. Koba, H. B. Nielsen, P. Olesen, Nucl. Phys. B 40, 317 (1972).
[19] L.D. Landau and E.M. Lifshitz. Statistical Physics (Course of Theoretical Physics, Volume 5),
Pergamon Press, Oxford, 1980.
[20] M.I. Gorenstein, Yad. Fiz. 31, 1630 (1980) (Russ.); Sov. J. Nucl. Phys. 31, 845 (1980).
[21] Weisstein, Eric W. ”Generalized Hypergeometric Function.” From MathWorld – A Wolfram Web
Resource. http://mathworld.wolfram.com/GeneralizedHypergeometricFunction.html
[22] A.P. Prudnikov, Yu.A. Brychkov, and O.I. Marichev, Integrals and Series, (Moscow, Nauka,
1986).
23
Recommended